In derivation process for luma motion vectors for merge mode, pruning of IvDC with A1 and B1 is incosistent between HTM and 3D-HEVC text. In HTM, if A1 or B1 is identical with IVMC (or T), it is marked as unavailable and is not used for pruning with IvDC. However, in the 3D-HEVC text, even in this case, A1 or B1 is still used for purning with IvDC, which is not reasonable and should be aligned with HTM.

In 3D-HEVC text, when no training samples are available for training IC parameters, IC is not performed, this is reasonable. However, in HTM, meaningless IC parameters are used in this case.

Should align HTM with 3D-HEVC text.

In the current WD text, I only found the following limitation for the prediction direction of DBBP. I.7.4.9.6 It is a requirement of bitstream conformance that, when dbbp_flag[ x0 ][ y0 ] is equal to 1, inter_pred_idc[ x0 ][ y0 ] shall not be equal to PRED_BI. However, the HTM has forced any inherited bi-prediction MV in DBBP to be forward prediction. I think, a bug-fix should be added on I.8.5.3.2.1, point 14 as following

14. When predFlagL0 is equal to 1 and predFlagL1 is equal to 1, and ( nOrigPbW + nOrigPbH ) is equal to 12 or dbbp_flag is equal to 1, the following applies:

refIdxL1 = −1 predFlagL1 = 0